A Gradient-Free and Parallel Hierarchical Motion Planning Framework for Quadrotor Swarm

Quadrotor swarm with advanced coordination capabilities has garnered widespread attention, yet efficient and reliable motion planning remains a challenge. This paper presents a hierarchical motion planning framework for quadrotor swarm autonomous navigation in unknown, cluttered scenes. Specifically, we take a step forward sampling of multiple neighbors in the lattice graph and generate a group of paths. The infeasible paths are excluded and the one with the minimum cost is chosen from the remaining. Taking it as a guiding path, a gradient-free trajectory optimization method based on model predictive path integral (MPPI) is developed to produce the execution trajectory. Compared to gradient-descent approaches, it is capable of dealing with non-continuous and non-convex constraints. Additionally, the proposed method is deployed on GPUs in parallel to enhance efficiency. Extensive simulations demonstrate the robustness and effectiveness of the proposed motion planning framework.